jskdr / keraspp Goto Github PK

View Code? Open in Web Editor NEW

177.0 20.0 96.0 6.87 MB

코딩셰프의 3분 딥러닝, 케라스맛

License: MIT License

Python 3.24% Jupyter Notebook 96.76%

ann dnn cnn rnn ae gan unet keras tensorflow

keraspp's People

Contributors

Stargazers

Watchers

Forkers

mwasa oppa3109 ohahohah duke24k jeilove daeukkim99 mkhoin yskeem runningmachine kayoungban lsj1888 pablitocho eggpan95 skim0119 jungi21cc clianor bigsea00001 astonisher emotionalcode eyejoy shubhampachori12110095 tkddus121 ip102803 rosssong grseb9s ldbslovis junn279 bearrundr hyukkyulee deesatzed kimpanjung meganad parandurume sskimin4 sjang42 ki815kaisian hubaimaster ayo79 charlie13 gipark2001 rheehot kim-sunghoon substage kwhkim daehongkim1 hzhe18 hero-kwon hellobwjung andrewjyjang lepuz pegasuskim-dev myeonghahwang tjddnjs0416 jieunchae seokleeus srshin dongshik juyongdo drworrier mapogentleman joosm harimjung keon bumsaengg hehpollon dreamerkim 0r0i yongsukki kidmam hgkim2004 primespace lovemewithoutall kimwooglae dinoforcered yeonhooy yangwonki pnustats jzzyuu gipallee devdio dontdevoy zerostone-kr robertchoi nzmars hanyangjack yeachan-heo kiseonjeong osokjs ccchai1 scentscnet holystar97 josungbork ljungee yjo731 for2000ad benjamincho90

keraspp's Issues

ex7_2_gan_cnn_mnist_tf.py를 실행하니 다음과 같은 에러가 납니다.

Using TensorFlow backend.
<function image_data_format at 0x7fb6200a4d90>
Output_fold is GAN_OUT
2018-11-25 08:56:45.589349: I tensorflow/core/platform/cpu_feature_guard.cc:141] Your CPU supports instructions that this TensorFlow binary was not compiled to
use: SSE4.1 SSE4.2 AVX
2018-11-25 08:56:45.679946: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:964] successful NUMA node read from SysFS had negative value (-1), but there
must be at least one NUMA node, so returning NUMA node zero
2018-11-25 08:56:45.680395: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1411] Found device 0 with properties:
name: Tesla K80 major: 3 minor: 7 memoryClockRate(GHz): 0.8235
pciBusID: 0000:00:04.0
totalMemory: 11.17GiB freeMemory: 11.09GiB
2018-11-25 08:56:45.680431: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1490] Adding visible gpu devices: 0
2018-11-25 08:56:46.007533: I tensorflow/core/common_runtime/gpu/gpu_device.cc:971] Device interconnect StreamExecutor with strength 1 edge matrix:
2018-11-25 08:56:46.007606: I tensorflow/core/common_runtime/gpu/gpu_device.cc:977] 0
2018-11-25 08:56:46.007617: I tensorflow/core/common_runtime/gpu/gpu_device.cc:990] 0: N
2018-11-25 08:56:46.007899: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1103] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 wi
th 10749 MB memory) -> physical GPU (device: 0, name: Tesla K80, pci bus id: 0000:00:04.0, compute capability: 3.7)
Epoch is 0
Number of batches 2
Traceback (most recent call last):
File "ex7_2_gan_cnn_mnist_tf.py", line 211, in
main()
File "ex7_2_gan_cnn_mnist_tf.py", line 207, in main
train(args)
File "ex7_2_gan_cnn_mnist_tf.py", line 166, in train
d_loss, g_loss = gan.train_both(x)
File "ex7_2_gan_cnn_mnist_tf.py", line 97, in train_both
g_loss = self.train_on_batch(z, [1] * ln)
File "/home/zeroone_gcp/anaconda3/envs/env-py36-gym-tf/lib/python3.6/site-packages/keras/engine/training.py", line 1211, in train_on_batch
class_weight=class_weight)
File "/home/zeroone_gcp/anaconda3/envs/env-py36-gym-tf/lib/python3.6/site-packages/keras/engine/training.py", line 677, in _standardize_user_data
self._set_inputs(x)
File "/home/zeroone_gcp/anaconda3/envs/env-py36-gym-tf/lib/python3.6/site-packages/keras/engine/training.py", line 636, in _set_inputs
outputs = self.call(unpack_singleton(self.inputs))
File "/home/zeroone_gcp/anaconda3/envs/env-py36-gym-tf/lib/python3.6/site-packages/keras/engine/network.py", line 561, in call
if cache_key in self._output_tensor_cache:
AttributeError: 'GAN' object has no attribute '_output_tensor_cache'

환경은 다음과 같습니다. (conda list)

packages in environment at

데이타 정규화를 하는데 5로 나눈 이유는 무엇인가?

    # data normalize
    data_dn = (data - np.mean(data)) / np.std(data) / 5
    plt.plot(data_dn)

data_dn = (data - np.mean(data)) / np.std(data)
이것만으로도 충분하지 않을까요?
5로 나눈 이유를 알고 싶습니다

Conv2D 사이즈 관련 질문드립니다

keraspp/ex6_2_ae_conv_mnist_mc.py

Line 29 in 4090fcc

z = Conv2D(1, (7, 7))(x)

리포지토리 코드와 책에 따르면 인코딩 모델의 마지막 레이어는 7x7 사이즈를 갖도록 설정되었는데요,

_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
input_12 (InputLayer)        (None, 28, 28, 1)         0         
_________________________________________________________________
cnn1 (Conv2D)                (None, 26, 26, 4)         40        
_________________________________________________________________
max_pooling1 (MaxPooling2D)  (None, 13, 13, 4)         0         
_________________________________________________________________
cnn3 (Conv2D)                (None, 11, 11, 8)         296       
_________________________________________________________________
max_pooling2 (MaxPooling2D)  (None, 6, 6, 8)           0         # <--- 이 부분
_________________________________________________________________
final_cnn (Conv2D)           (None, 1, 1, 1)           289       
=================================================================
Total params: 625
Trainable params: 625
Non-trainable params: 0
_________________________________________________________________

max_pooling2 레이어의 출력값이 6x6 인데 CNN 7x7 을 사용할 수 있나요?

위 summary() 결과를 얻기 위해 사용한 모델 코드는 아래와 같습니다.

        input_layer = Input(shape=self.input_shape)
        
        x = Conv2D(4, (3, 3), name='cnn1')(input_layer)
        x = MaxPooling2D((2, 2), padding='same', name='max_pooling1')(x)
        x = Conv2D(8, (3, 3), name='cnn3')(x)
        x = MaxPooling2D((2, 2), padding='same', name='max_pooling2')(x)
        
        output_layer = Conv2D(1, (6, 6), name='final_cnn')(x)

또한 7x7 사이즈의 CNN 을 사용했을 때, 나오는 에러는 아래와 같습니다.

ValueError: Negative dimension size caused by subtracting 7 from 6 for 'final_cnn_4/convolution' (op: 'Conv2D') with input shapes: [?,6,6,8], [7,7,8,1].

ANN/DNN에서 weight와 bias값들을 출력 할 수 있나요?

제가 keras에서 구현한 ANN 혹은 DNN 학습 결과를 C/C++로 된 프로그램에 적용하고자 합니다.
그러기 위해서 C/C++로 Neural network 코드를 만들었고, keras에서 계산한 hidden layer와 출력 layer의
weight와 bias 값을 출력해서 C/C++ 코드에 적용하려고 합니다.
가능할까요? 부탁 드립니다!

ex6_1_ae를 cpu 텐서플로에서 돌리면 epochs를 늘여도 수렴이 잘 안됩니다.

파이썬3.7, 텐서플로2.0 cpu 버전입니다.
에포크에 따른 accuracy 변화는 아래와 같습니다.
책에 나온 결과와 차이가 매우 크게 납니다.

왜 그런 걸까요.

runfile('D:/Python/merong.py', wdir='D:/Python')
(60000, 784)
(10000, 784)
Train on 60000 samples, validate on 10000 samples
Epoch 1/10
60000/60000 [==============================] - 7s 123us/sample - loss: 0.6959 - accuracy: 0.3923 - val_loss: 0.6958 - val_accuracy: 0.3924
Epoch 2/10
60000/60000 [==============================] - 5s 84us/sample - loss: 0.6956 - accuracy: 0.3941 - val_loss: 0.6956 - val_accuracy: 0.3943
Epoch 3/10
60000/60000 [==============================] - 6s 108us/sample - loss: 0.6954 - accuracy: 0.3959 - val_loss: 0.6954 - val_accuracy: 0.3960
Epoch 4/10
60000/60000 [==============================] - 6s 104us/sample - loss: 0.6952 - accuracy: 0.3978 - val_loss: 0.6952 - val_accuracy: 0.3979
Epoch 5/10
60000/60000 [==============================] - 5s 86us/sample - loss: 0.6950 - accuracy: 0.3995 - val_loss: 0.6950 - val_accuracy: 0.3996
Epoch 6/10
60000/60000 [==============================] - 6s 101us/sample - loss: 0.6948 - accuracy: 0.4013 - val_loss: 0.6948 - val_accuracy: 0.4014
Epoch 7/10
60000/60000 [==============================] - 5s 87us/sample - loss: 0.6946 - accuracy: 0.4031 - val_loss: 0.6946 - val_accuracy: 0.4032
Epoch 8/10
60000/60000 [==============================] - 5s 84us/sample - loss: 0.6944 - accuracy: 0.4048 - val_loss: 0.6944 - val_accuracy: 0.4049
Epoch 9/10
60000/60000 [==============================] - 8s 134us/sample - loss: 0.6942 - accuracy: 0.4066 - val_loss: 0.6942 - val_accuracy: 0.4067
Epoch 10/10
60000/60000 [==============================] - 5s 90us/sample - loss: 0.6940 - accuracy: 0.4083 - val_loss: 0.6940 - val_accuracy: 0.4084

ex9_3_advanced_keras.py 에서 발생하는 ValueError

Backend_for_Lambda()의 yp = m.predict_on_batch([[1, 2, 3], [3, 4, 8]]) 부분에서 다음과 같은 에러가 발생하고 있습니다.

ValueError: Error when checking model input: the list of Numpy arrays that you are passing to your model is not the size the model expected. Expected to see 1 array(s), but instead got the following list of 2 arrays: [array([[1],
       [2],
       [3]]), array([[3],
       [4],
       [8]])]...

https://keras.io/ko/models/sequential/의 predict_on_batch를 참고해 보았지만, 어떻게 해결하면 좋을지 문의하고자 이슈 남깁니다.

ex2_1_ann_mnist_cl.py에서 plot_acc()가 작동하지 않는 문제

ex4_2_cnn_cifar10_cl 실행하니 you forgot to call super(YourClass, self).init() 에러가 납니다.

ex4_2_cnn_cifar10_cl 실행하니 아래와 같은 에러가 납니다.
도와주세요.

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "C:/Users/softg/PycharmProjects/test/keraspp/ex4_2_cnn_cifar10_cl.py", line 22, in <module>
    main()
  File "C:/Users/softg/PycharmProjects/test/keraspp/ex4_2_cnn_cifar10_cl.py", line 18, in main
    m = Machine()
  File "C:/Users/softg/PycharmProjects/test/keraspp/ex4_2_cnn_cifar10_cl.py", line 14, in __init__
    super().__init__(X, y, nb_classes=10)
  File "C:\Users\softg\PycharmProjects\test\keraspp\keraspp\aicnn.py", line 125, in __init__
    self.set_model()
  File "C:\Users\softg\PycharmProjects\test\keraspp\keraspp\aicnn.py", line 136, in set_model
    self.model = CNN(nb_classes=nb_classes, in_shape=data.input_shape)
  File "C:\Users\softg\PycharmProjects\test\keraspp\keraspp\aicnn.py", line 25, in __init__
    model.build_model()
  File "C:\Users\softg\PycharmProjects\test\keraspp\keraspp\aicnn.py", line 49, in build_model
    model.cl_part = Model(x, z_cl)
  File "C:\Users\softg\AppData\Local\Programs\Python\Python36\lib\site-packages\keras\engine\network.py", line 316, in __setattr__
    'It looks like you are subclassing `Model` and you '
RuntimeError: It looks like you are subclassing `Model` and you forgot to call `super(YourClass, self).__init__()`. Always start with this line.

Process finished with exit code 1

jskdr / keraspp Goto Github PK

keraspp's People

Contributors

Stargazers

Watchers

Forkers

keraspp's Issues

ex7_2_gan_cnn_mnist_tf.py를 실행하니 다음과 같은 에러가 납니다.

packages in environment at

Name Version Build Channel

데이타 정규화를 하는데 5로 나눈 이유는 무엇인가?

Conv2D 사이즈 관련 질문드립니다

ANN/DNN에서 weight와 bias값들을 출력 할 수 있나요?

ex6_1_ae를 cpu 텐서플로에서 돌리면 epochs를 늘여도 수렴이 잘 안됩니다.

ex9_3_advanced_keras.py 에서 발생하는 ValueError

ex2_1_ann_mnist_cl.py에서 plot_acc()가 작동하지 않는 문제

ex4_2_cnn_cifar10_cl 실행하니 you forgot to call super(YourClass, self).init() 에러가 납니다.

Recommend Projects

React

Vue.js

Typescript

TensorFlow

Django

Laravel

D3

Recommend Topics

javascript

web

server

Machine learning

Visualization

Game

Recommend Org

Facebook

Microsoft

Google

Alibaba

D3

Tencent

Jobs